From patchwork Fri Sep 28 16:15:02 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Lam, Tiago" X-Patchwork-Id: 976382 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (mailfrom) smtp.mailfrom=openvswitch.org (client-ip=140.211.169.12; helo=mail.linuxfoundation.org; envelope-from=ovs-dev-bounces@openvswitch.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=intel.com Received: from mail.linuxfoundation.org (mail.linuxfoundation.org [140.211.169.12]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 42MGxq3WQTz9s5c for ; Sat, 29 Sep 2018 02:15:59 +1000 (AEST) Received: from mail.linux-foundation.org (localhost [127.0.0.1]) by mail.linuxfoundation.org (Postfix) with ESMTP id 5BE6FD95; Fri, 28 Sep 2018 16:15:27 +0000 (UTC) X-Original-To: ovs-dev@openvswitch.org Delivered-To: ovs-dev@mail.linuxfoundation.org Received: from smtp1.linuxfoundation.org (smtp1.linux-foundation.org [172.17.192.35]) by mail.linuxfoundation.org (Postfix) with ESMTPS id 0EC8FD77 for ; Fri, 28 Sep 2018 16:15:25 +0000 (UTC) X-Greylist: domain auto-whitelisted by SQLgrey-1.7.6 Received: from mga07.intel.com (mga07.intel.com [134.134.136.100]) by smtp1.linuxfoundation.org (Postfix) with ESMTPS id 62F9B7E2 for ; Fri, 28 Sep 2018 16:15:24 +0000 (UTC) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga005.fm.intel.com ([10.253.24.32]) by orsmga105.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 28 Sep 2018 09:15:24 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.54,315,1534834800"; d="scan'208";a="266810227" Received: from silpixa00399125.ir.intel.com ([10.237.223.34]) by fmsmga005.fm.intel.com with ESMTP; 28 Sep 2018 09:15:22 -0700 From: Tiago Lam To: ovs-dev@openvswitch.org Date: Fri, 28 Sep 2018 17:15:02 +0100 Message-Id: <1538151315-117132-2-git-send-email-tiago.lam@intel.com> X-Mailer: git-send-email 2.7.4 In-Reply-To: <1538151315-117132-1-git-send-email-tiago.lam@intel.com> References: <1538151315-117132-1-git-send-email-tiago.lam@intel.com> X-Spam-Status: No, score=-4.2 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_MED autolearn=ham version=3.3.1 X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on smtp1.linux-foundation.org Cc: fbl@sysclose.org, i.maximets@samsung.com Subject: [ovs-dev] [PATCH v10 01/14] netdev-dpdk: fix mbuf sizing X-BeenThere: ovs-dev@openvswitch.org X-Mailman-Version: 2.1.12 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , MIME-Version: 1.0 Sender: ovs-dev-bounces@openvswitch.org Errors-To: ovs-dev-bounces@openvswitch.org From: Mark Kavanagh There are numerous factors that must be considered when calculating the size of an mbuf: - the data portion of the mbuf must be sized in accordance With Rx buffer alignment (typically 1024B). So, for example, in order to successfully receive and capture a 1500B packet, mbufs with a data portion of size 2048B must be used. - in OvS, the elements that comprise an mbuf are: * the dp packet, which includes a struct rte mbuf (704B) * RTE_PKTMBUF_HEADROOM (128B) * packet data (aligned to 1k, as previously described) * RTE_PKTMBUF_TAILROOM (typically 0) Some PMDs require that the total mbuf size (i.e. the total sum of all of the above-listed components' lengths) is cache-aligned. To satisfy this requirement, it may be necessary to round up the total mbuf size with respect to cacheline size. In doing so, it's possible that the dp_packet's data portion is inadvertently increased in size, such that it no longer adheres to Rx buffer alignment. Consequently, the following property of the mbuf no longer holds true: mbuf.data_len == mbuf.buf_len - mbuf.data_off This creates a problem in the case of multi-segment mbufs, where that assumption is assumed to be true for all but the final segment in an mbuf chain. Resolve this issue by adjusting the size of the mbuf's private data portion, as opposed to the packet data portion when aligning mbuf size to cachelines. Fixes: 4be4d22 ("netdev-dpdk: clean up mbuf initialization") Fixes: 31b88c9 ("netdev-dpdk: round up mbuf_size to cache_line_size") CC: Santosh Shukla Signed-off-by: Mark Kavanagh Acked-by: Santosh Shukla Acked-by: Eelco Chaudron --- lib/netdev-dpdk.c | 56 +++++++++++++++++++++++++++++++++++++------------------ 1 file changed, 38 insertions(+), 18 deletions(-) diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c index f91aa27..1e19622 100644 --- a/lib/netdev-dpdk.c +++ b/lib/netdev-dpdk.c @@ -88,10 +88,6 @@ static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); #define MTU_TO_MAX_FRAME_LEN(mtu) ((mtu) + ETHER_HDR_MAX_LEN) #define FRAME_LEN_TO_MTU(frame_len) ((frame_len) \ - ETHER_HDR_LEN - ETHER_CRC_LEN) -#define MBUF_SIZE(mtu) ROUND_UP((MTU_TO_MAX_FRAME_LEN(mtu) \ - + sizeof(struct dp_packet) \ - + RTE_PKTMBUF_HEADROOM), \ - RTE_CACHE_LINE_SIZE) #define NETDEV_DPDK_MBUF_ALIGN 1024 #define NETDEV_DPDK_MAX_PKT_LEN 9728 @@ -637,7 +633,11 @@ dpdk_mp_create(struct netdev_dpdk *dev, int mtu, bool per_port_mp) char mp_name[RTE_MEMPOOL_NAMESIZE]; const char *netdev_name = netdev_get_name(&dev->up); int socket_id = dev->requested_socket_id; - uint32_t n_mbufs; + uint32_t n_mbufs = 0; + uint32_t mbuf_size = 0; + uint32_t aligned_mbuf_size = 0; + uint32_t mbuf_priv_data_len = 0; + uint32_t pkt_size = 0; uint32_t hash = hash_string(netdev_name, 0); struct dpdk_mp *dmp = NULL; int ret; @@ -650,6 +650,9 @@ dpdk_mp_create(struct netdev_dpdk *dev, int mtu, bool per_port_mp) dmp->mtu = mtu; dmp->refcount = 1; + /* Get the size of each mbuf, based on the MTU */ + mbuf_size = dpdk_buf_size(dev->requested_mtu); + n_mbufs = dpdk_calculate_mbufs(dev, mtu, per_port_mp); do { @@ -661,8 +664,8 @@ dpdk_mp_create(struct netdev_dpdk *dev, int mtu, bool per_port_mp) * so this is not an issue for tasks such as debugging. */ ret = snprintf(mp_name, RTE_MEMPOOL_NAMESIZE, - "ovs%08x%02d%05d%07u", - hash, socket_id, mtu, n_mbufs); + "ovs%08x%02d%05d%07u", + hash, socket_id, mtu, n_mbufs); if (ret < 0 || ret >= RTE_MEMPOOL_NAMESIZE) { VLOG_DBG("snprintf returned %d. " "Failed to generate a mempool name for \"%s\". " @@ -671,17 +674,34 @@ dpdk_mp_create(struct netdev_dpdk *dev, int mtu, bool per_port_mp) break; } - VLOG_DBG("Port %s: Requesting a mempool of %u mbufs " - "on socket %d for %d Rx and %d Tx queues.", - netdev_name, n_mbufs, socket_id, - dev->requested_n_rxq, dev->requested_n_txq); - - dmp->mp = rte_pktmbuf_pool_create(mp_name, n_mbufs, - MP_CACHE_SZ, - sizeof (struct dp_packet) - - sizeof (struct rte_mbuf), - MBUF_SIZE(mtu) - - sizeof(struct dp_packet), + VLOG_DBG("Port %s: Requesting a mempool of %u mbufs of size %u " + "on socket %d for %d Rx and %d Tx queues, " + "cache line size of %u", + netdev_name, n_mbufs, mbuf_size, socket_id, + dev->requested_n_rxq, dev->requested_n_txq, + RTE_CACHE_LINE_SIZE); + + mbuf_priv_data_len = sizeof(struct dp_packet) - + sizeof(struct rte_mbuf); + /* The size of the entire dp_packet. */ + pkt_size = sizeof (struct dp_packet) + + mbuf_size + RTE_PKTMBUF_HEADROOM; + /* mbuf size, rounded up to cacheline size. */ + aligned_mbuf_size = ROUND_UP(pkt_size, RTE_CACHE_LINE_SIZE); + /* If there is a size discrepancy, add padding to mbuf_priv_data_len. + * This maintains mbuf size cache alignment, while also honoring RX + * buffer alignment in the data portion of the mbuf. If this adjustment + * is not made, there is a possiblity later on that for an element of + * the mempool, buf, buf->data_len < (buf->buf_len - buf->data_off). + * This is problematic in the case of multi-segment mbufs, particularly + * when an mbuf segment needs to be resized (when [push|popp]ing a VLAN + * header, for example. + */ + mbuf_priv_data_len += (aligned_mbuf_size - pkt_size); + + dmp->mp = rte_pktmbuf_pool_create(mp_name, n_mbufs, MP_CACHE_SZ, + mbuf_priv_data_len, + mbuf_size + RTE_PKTMBUF_HEADROOM, socket_id); if (dmp->mp) {